Representing Minds

by
Robert M. Gordon
gordon@umsl.edu

[Paper presented at Tucson 2002 (Toward a Science of Consciousness), Tucson Arizona April 8th 2002]

Here are quotations from two philosophers, one of them perhaps the most important of the last two millennia; the other, among the most important of the last century:

1. It is obvious that if one wants to represent a thinking being, one must put oneself in its place, and thus substitute one's own subject for the object one wants to consider (which is not the case in any other species of investigation).^{1}
2. We project ourselves into what, from his remarks and other indications, we imagine the speaker's state of mind to have been, . . . even into what from his behavior we imagine a mouse's state of mind to have been, and dramatize it as a belief, wish or striving, verbalized as seems relevant and natural to us in the state thus feigned^{2}

I'll use the term role-taking simulation to cover both Kant's "putting myself in his place" and Quine's "projecting ourselves into what we imagine another's state of mind to be." The main point of calling this role-taking simulation is to distinguish it from the various subpersonal processes by which we automatically mirror the minds of other human beings even without an imaginative leap into their shoes: processes that predispose us to attend to what others are attending to, for example, by following their direction of gaze; or predispose us to react as they do, for example, by facial mimicry and the somatosensory responses that seem important for recognizing the emotions of others; or predispose us to intend what they intend, by way of mirror neurons.^{3}

According to both Kant and Quine, an act of role-taking simulation is essential to representing something as having mental states: for Kant, doing so mentally, for Quine, doing so verbally. It is important to notice that in their view role-taking simulation is not a heuristic for discovering something that is independently conceivable. The main point is not that you have to use such a methodology to discover the contents of other minds, but that you have to do this even to understand that there are minds. If you are to grasp the idea that there are subjects for whom things are thus-and-so, you can only do this by putting yourself in their place, according to Kant. Talk of beliefs, desires, fears, and other propositional attitudes, says Quine, is an essentially dramatic idiom. There is no option of stepping outside one's simulation—of taking off the distorting prisms of imagination in order to check and see "objectively" what state of mind a person actually is in. Take off the prisms and it all goes away—we see nothing out there but a mechanism. (For both philosophers the message is cautionary. Kant warned his readers that this "obvious" truth about the mere form of our representation of thinking beings gives us no warrant for a dualistic ontology. For Quine, to say that without the prisms we see nothing there but a mechanism is to say that we see nothing more than what is really there.)^{4}

There is a small problem and a large problem with the views I am attributing to Kant and Quine. First, what I think is only a small problem: There is simply no reason to think that whenever we try to understand, explain, or predict another's behavior we are engaging in role-taking simulation. All the more, there is no reason to suppose we engage in this activity every time we conceive or speak of something as a thinking being, or whenever we treat something as a thinking being. That problem, I believe, is easily fixed. What should be said is this: We conceive and treat the object as a placeholder for such an operation. Roughly, to conceive or speak of an object as a thinking being, a mentally endowed being, is to conceive or speak of it as an object to which we might properly mentally transport ourselves.^{5} It follows that people who have thought of volcanoes, fire, and the sun as thinking or mind-endowed have also regarded them as proper objects for role-taking simulation. Many of us think playfully, not seriously, of automobiles and personal computers as mind-endowed—usually, as we say, endowed with minds of their own. And so we also think of these playfully, not seriously, as proper objects for role-taking simulation. Of course, this would have to be simulation manqué: for one thing, it's hard to pick out their sensors and effectors, or to invoke the automatic mimicry that shapes our responses to human faces and bodies, and to a lesser degree to those of other animals.

Role-taking simulation is not enough

The bigger problem is that conceiving something as a proper object for role-taking simulation is clearly not sufficient for representing something as a thinking being, as having mental states. One reason concerns the developmental gap between the two. Children as young as age two are capable of engaging in pretend decision-making in the role of a particular person, real or fictional. They can also recognize when another person is engaged in such play. But it is not until about two years later, around age four, that children exhibit an explicit understanding of people as thinking or mind-endowed entities—a point I will return to at the end.

This is not to say that younger children cannot explain or anticipate behavior by using simulation. For explaining and predicting a great deal of the human behavior they are likely to encounter in real life or in stories, they can probably get by with a relatively simple employment of their imaginative abilities. For example: Sally and Anne are playing together. Anne takes a marble out of Sally's box, drops it into her own basket, and then goes out of the room. Now Sally wants to play with her marble. Where does she look for it? Taking the role of Sally in this simple scenario, a child should have no trouble deciding where to look for her marble: in Anne's basket. One might reason:

However, suppose that in the story Sally is away when Anne takes her marble. For example: Sally and Anne are playing together. Sally goes out of the room. Anne takes a marble out of Sally's box and drops it into her own basket. Sally comes back. She wants to play with her marble. Where does she look for it? Here we don't want to reason to the conclusion,

Rather, we want,

At the same time, however, we want to anticipate Sally's disappointment when she fails to find her marble where she left it. This requires that we maintain our "objectivity." So, this is what we have to pretend:

That seems unremarkable enough, but we are talking about taking the role of Sally. So it would seem that what must be pretended is the following:

1. (a) Sally's marble has been moved to Anne's basket, but (b) I don't believe it has.

Before we consider how to pretend this, let's notice that the logical form of M is quite special. It is,

This form has attracted a lot of interest among 20^th century philosophers. G. E. Moore noted that, even though sentences of this form are logically consistent, to utter them as statements or assertions is, in some pragmatic way, inconsistent. This is sometimes referred to as Moore's Paradox, and the sentences or statements that instantiate it are sometimes referred to as Moorean absurdities. According to Moore, when I say "Sally's marble has been moved to Anne's basket," I imply that I believe that Sally's marble has been moved to Anne's basket. There is no contradiction, because in asserting (Ma) I only imply that I believe (Ma); I don't assert that I do.

There are several different accounts of the inconsistency of Moorean absurdities.^{6} I will not take a stand on the general issue. However, where pretending

is concerned, it seems clear that one source of inconsistency is this: If within a pretense, p is true, then within the pretense, the "fact" that p is accessible to reasoning, including reasoning on the basis of which I (as myself or in a role) act.^{7} However, if p is accessible to reasoning on the basis of which I act, then I believe that p. If I do not believe that p, then the fact or pretend fact that p is not a reason of mine, or a part of a reason of mine, nor even a candidate for such a status. So, within the pretense a Moorean absurdity would introduce a conflict between truth and accessibility.

The problem with pretending M, for example, is that if we are to pretend (Ma), then we must allow (Ma) to become a premise of our reasoning concerning where the marble is now. However, if we are to pretend (Mb), then we must not allow (Ma) to become a premise of our—that is, Sally's—reasoning concerning where the marble is now.. Within the role of Sally, we must pretend ignorance of (Ma). This will require that our pretense be split or compartmentalized in way that allows both truth and inaccessibility.

To better illustrate the point, consider another Moorean absurdity:

1. (a) This mushroom is poisonous, but (b) I do not believe it is.

You can indeed suppose—or pretend or imagine—that a certain mushroom is poisonous and that you are ignorant of this fact: You picture yourself innocently sautéing your poisonous false morel, then eating it and getting very sick. This requires, however, that the pretend-player be capable of engaging in a compartmentalized double pretense. You feed contrary, incompatible premises into the two compartments. Into one compartment you feed the premise,

Into the other, you feed the contrary premise,

One pretense has access to Ma, and the other has access to Ma'. Neither compartment knows what is going on in the other. Each is informationally encapsulated.

But compartmentalization isn't enough. If in one compartment you act on the basis of Pa and in the other you act on the basis of Pa', then you are just pretending to be split-minded—perhaps engaging in a tug-of-war with yourself, eating the mushroom with one hand and pulling it away from your mouth with the other. To avoid this absurd behavior, it is crucial that only one compartment be dedicated to motivating the action—I will call that the inner pretense—and the other to describing and predicting the actual circumstances and effects of the action—the outer pretense. On the premise that the mushroom is a real morel, you eat it; on the premise that it is a false morel, you begin to act sick. The inner pretense takes Pa' as input. Your reasoning might go: This is a genuine morel. If it is, then it is not poisonous, and it would be fabulous to eat sautéed in butter. So I'll pick the mushroom, take it home, sauté it, and enjoy.

The outer pretense, on the other hand, takes the contrary premise Pa as input. Adding Pa to your background knowledge, you reason, OK, false morel, this is poisonous, maybe not lethal, but bad. Given the poisonous properties of the mushroom, you are nomologically bound, in this closest possible world in which the mushroom is a false morel, to be poisoned. Playing the role, you display in your behavior the operations of unseen forces: You begin to roll on the ground and moan in feigned sickness. Only then do the two pretenses, the inner and the outer, merge: It's all out in the open now, the knot is untied, bringing the dramatic irony to an end. You realize you've been hoodwinked by Mother Nature. You, or the person you are playing, were unaware of a fact that significantly affected the outcome of your behavior.

Let's return to the original Sally-Anne story. To deal with the case in which Anne moves the marble after Sally has left the room, we had to pretend a Moorean absurdity, namely,

1. (a) Sally's marble has been moved to Anne's basket, but (b) I don't believe it has.

To do this we compartmentalize our pretense, restricting the premise,

to the outer compartment. Because the news of (Ma) isn't allowed to enter the inner compartment, our practical reasoning relies on the default assumption,

So we might reason:

Implicit Belief Ascription

I said that role-taking simulation is not sufficient for representing something as a thinking being. Not all people who engage in it exhibit an explicit understanding of people as thinking or mind-endowed entities; there is a developmental gap between these competences. However, we can make good sense of the claim that in all role-taking simulation one is at least implicitly ascribing mental states. For example, in the simple Sally-Anne scenario, where Anne took her marble while Sally was watching, we reasoned in the role of Sally:

we implicitly ascribe the corresponding belief to Sally. For we treat that premise as accessible to Sally's practical reasoning, as at least potentially, and in this case actually, a part of the basis on which she acts. That requires that she believe it.^{8} This implicit ascription does not depend on possession of the concept of belief or of knowledge. For it does not require explicitly pretending two things, one about this object, the marble, and the other about oneself, in the role of Sally: that the marble has been moved, and that I, Sally, am aware of that, and therefore believe it to be so. Unless we take the step of relegating (Ma) to an outer pretense, our procedure makes it accessible to Sally's reasoning. Thus it excludes from our pretense all possible worlds in which it is the case that (Ma) and yet I (in the role of Sally) do not believe that (Ma). In that sense, we might say that a belief ascription is procedurally implicit in role-taking simulation. Within the simulation, Sally's beliefs march in lockstep with the facts, just as if we were ascribing knowledge or awareness of the facts.

Explicit Belief Ascription

What makes explicit belief ascription possible is the capacity to separate factuality or truth within a pretense from accessibility to reasoning that leads to action. This requires competence in carrying out a compartmentalized double pretense in which different and even contrary premises may be introduced into two distinct lines of reasoning: so that, for example, "objectively," the marble has been moved to Anne's basket and therefore can be found there, not in Sally's box; whereas "subjectively"—that is, for the purpose of deciding what to do in the role of Sally—it hasn't been moved from Sally's box, and therefore can still be found there. The compartmentalization of pretense makes belief and truth orthogonal, opening up the following four possibilities:

Interactionist Compartmentalization

I have distinguished between compartmentalized and uncompartmentalized pretense. But strictly speaking, all pretend play requires compartmentalization of a sort. If a child simply pretends that the banana is a telephone and proceeds to use it as a telephone, the child is acting on the basis of pretend facts, and therefore pretend beliefs. The pretend belief that the banana is a telephone, together with the various ramifications this belief introduces into the pretend world, are counterfactual beliefs. Therefore, they should be quarantined or kept separate from those beliefs that guide the child's action in the actual world. We should hope, for example, that an extremely hungry child, a starving child, would not hesitate to eat the banana-telephone. In these cases, only one of the two compartments is a pretense; the idea is to cordon off the pretense from one's action-guiding reasoning—to run the decision-making off-line, or to allow it on-line only within strict limits.

However, the pretense required for pretending a Moorean absurdity,

requires a different kind of compartmentalization. In the false morel story, for example, we do want separation between the two pretenses, but we also want interaction. The act of eating a particular genuine morel in the inner pretense is coextensional with the ingestion of a particular poisonous false morel in the outer pretense. The consequences of this unfortunate identity eventually infiltrate into the inner pretense, as you the protagonist find yourself pretend-sick and pretend-miserable. In the Sally-Anne story, the translocation of the marble in the outer pretense comes to invade the inner pretense only when you, as Sally, fail to find the marble in your box. These are examples of interactionist compartmentalization.

To be sure, even interactionist compartmentalization does not require that both compartments be pretenses. Suppose someone boasts to me that he knows mushrooms very well, though I know he doesn't; and I want to prove the point to him indelibly. I see him picking a false morel, and I say nothing; putting myself in his place, I plan to eat it with gusto, and so I anticipate his doing that; but also, remaining in my own place, I anticipate his consequent sickness. The latter, outer compartment is not pretense; it's the real world, the world according to me.

So perhaps we can conclude: What is required for conceiving an object as thinking or mind-endowed is to conceive it as placeholder for, not just role-taking simulation, but role-taking simulation with compartmentalization—and, more narrowly, interactionist compartmentalization, where there are identities that allow actions motivated within the inner compartment to make a difference in the outer compartment, where this difference may come back to affect the inner compartment in unanticipated ways.

Endnote: Beyond Belief?

I said that it isn't until about age four that children exhibit an explicit understanding of people as thinking or mind-endowed entities. However, the relevant experiments, such as the famous Sally-Ann and rock-sponge experiments, concern only the ascription of belief, knowledge, and the appearance-reality distinction. Well before age four, children recognize that others have desires different from their own. Aren't they therefore already representing people as having mental states?

One worry is this: Are the desires that younger children ascribe, whether verbally or in anticipating behavior, substantially different from the inner states of a homing rocket or a smart bomb? I mean a bomb or rocket that adjusts to changing conditions that might throw it off target. If they are only representing people as having inner states like these, then, putting aside the appropriateness of the term "mental," we can certainly agree that they would be missing something important that we think characteristic of mental states.

A part of what they would seem to be missing is intentionality, although again I don't want to put much weight on the terminology. I'll just give examples. Recall that pretending the Moorean absurdity,

requires interactionist compartmentalization. But so do the following pretenses, I believe:

If this is right, then we can conclude, without quibbling about terms, that until and unless one is capable of interactionist compartmentalization, one's representation of others is missing important elements of intentionality and mentality.

 Notes

1. Kant, I. 1781/1998. Critique of Pure Reason, first edition. Tr. Guyer and Wood, Second Paralogism of Pure Reason. Scholarly note: The passage quoted appears only in the First Edition of the Critique of Pure Reason, under the "Second Paralogism of Pure Reason." But in both editions we find the following:

Now I cannot have the least representation of a thinking being, through an external experience, but only through self-consciousness. Thus such objects are nothing further than the transference of this consciousness of mine to other things, which can be represented as thinking beings only in this way. (Kant 1781/1998: 347/405) [The term objects as used here should be interpreted as "objects of thought."] [back to text]

 

2. Word and Object, 1960, 218 -219. Although Quine does not say much to amplify these remarks in Word and Object, in his recent book, The Pursuit of Truth, he speaks at length of the essential role of empathy in language learning and translation. [back to text]

 

3. As I speak now, I am not aware of putting myself in anyone's place. Yet it is a plausible hypothesis that my brain is rehearsing before an internal audience at least the gist of what I am saying, before I say it—or in less fortunate cases, a fraction of a second afterward, usually in time for a mid-sentence correction. It is also plausible that this internal audience has been calibrated, accurately or not, to my actual audience. It is also plausible that when your brain analyzes my utterances it does so by invoking many of the processes used to synthesize your own utterances. So, listening to what I am saying from your point of view, I will probably not have to put myself in your place, unless I see an unanticipated expression on several faces in the audience, or a mass walkout. [back to text]

 

4. Unfortunately, that would arguably make people with autism perhaps the only true realists, the only people who base their everyday actions on a correct ontology. [back to text]

 

5. And when I say "mentally transport ourselves" I do not mean as ourselves. It isn't a matter of putting Robert Gordon in, say, Napoleon's shoes (a point I've discussed in a number of publications). [back to text]

 

6. See Hroeska-Hardy, unpublished. [back to text]

 

7. This is incorrectly phrased, to be sure, but convenient in a talk. [back to text]

 

8. Even that she knows it, that she is aware that Anne moved the marble. But I needn't argue the point here. [back to text]